BackgroundComorbidity is increasingly important in the medical literature, with ever-increasing implications for diagnosis, treatment, prognosis, management and health care. The objective of this study is to measure casual versus causal comorbidity in primary care in three family practice populations.MethodsThis is a longitudinal observational study using the Transition Project datasets. Transition Project family doctors in the Netherlands, Malta and Serbia recorded details of all patient contacts in an episode of care structure using electronic medical records and the International Classification of Primary Care, collecting data on all elements of the doctor-patient encounter, including diagnoses (1,178,178 in the Netherlands, 93,606 in Malta, 405,150 in Serbia), observing 158,370 patient years in the Netherlands, 43,577 in Malta, 72,673 in Serbia. Comorbidity was measured using the odds ratio of both conditions being incident or rest-prevalent in the same patient in one-year dataframes, as against not, corrected for the prior probability of such co-occurrence, between the 41 joint most prevalent (joint top 20) episode titles in the three populations. Specific associations were explored in different age groups to observe the changes in odds ratios with increasing age as a surrogate for a temporal or biological gradient.ResultsThe high frequency of observed comorbidity with low consistency in both clinically and statistically significant odds ratios across populations indicates more casual than causal associations. A causal relationship would be expected to be manifest more consistently across populations. Even in the minority of cases where odds ratios were consistent between countries and numerically larger, those associations were observed to weaken with increasing patient age.ConclusionAfter applying accepted criteria for testing the causality of associations, most observed primary care comorbidity is due to chance, likely as a result of increasing illness diversity.Trial registrationThis study was performed on electronic patient record datasets made publicly available by the University of Amsterdam Department of General Practice, and did not involve any patient intervention.